APK-Malware-Dataset(2025)

Published: 8 October 2026| Version 1 | DOI: 10.17632/gsxhtvhbvj.1
Contributors:
,
,
,

Description

This dataset presents a structured static-analysis feature representation of Android applications (APK files) designed for research in Android cybersecurity, malware detection, malware-family classification, permission analysis, behavioral characterization, and machine-learning-based threat analysis. The dataset contains the extracted characteristics of 450 Android APK samples, transformed from raw application packages into a structured numerical and categorical representation suitable for statistical analysis and machine-learning workflows. Instead of requiring researchers to repeatedly reverse-engineer each APK to obtain basic security characteristics, this dataset provides a consolidated representation of the analyzed applications in a single feature table.

Files

Steps to reproduce

## Data Collection and Reproducibility The dataset was created by applying a reproducible **static-analysis workflow to 450 Android APK samples**. The APKs were analyzed without executing them to extract security and application characteristics. For each APK, information was extracted from the **AndroidManifest.xml, DEX files, native libraries, and application structure**. The extracted features include Android permissions, application metadata, SDK versions, APK size, DEX/native-library characteristics, API calls, Android components, exported components, URLs/domains/IPs, network indicators, behavioral indicators, and malware/detection information. The workflow was implemented using **Python 3**, with tools and libraries including **Androguard, JADX, XML processing tools, Pandas, NumPy, and Scikit-learn**. Supplementary malware information was obtained using available detection sources such as **VirusTotal, MalwareBazaar, and ClamAV**, where applicable. Android permissions were converted into binary features, where **1 indicates that a permission is present and 0 indicates that it is absent**. The extracted information was then combined into a structured feature matrix. The final dataset contains **450 APK samples and 513 features** and is provided as `APK-Malware-Dataset(2025).csv`. Supporting files provide the feature-level metadata and dataset-level statistical characterization. The reproducible workflow is: **APK Collection → APK Decoding → Feature Extraction → Permission/API/Component Analysis → Network & Behavior Analysis → Malware Information → Feature Matrix → Data Validation** The resulting dataset can be used for **binary malware classification, multi-class malware-family classification, feature selection, dimensionality reduction, clustering, and multidimensional security analysis**.

Institutions

Categories

Cybersecurity, Data Analytics Cybersecurity

Licence